*Corresponding author. Current address: EMPA St. Gallen, Ler-chenfeldstrasse 5, CH-9014, St. Gallen, Switzerland. Tel.: #41-71-274-7695; fax: #41-71-274-7788.

E-mail address: arie.bruinink@empa.ch (A. Bruinink).Abbreviations: ALP: Alkaline phosphatase; RBMC: Rat bone marrowcells; PBS: Phosphate bu!ered saline; TRAP: Tartrate-resistant acidphosphatase

Biomaterials 22 (2001) 3169}3178

The sti!ness of bone marrow cell}knit composites is increasedduring mechanical load

A. Bruinink*, D. Siragusano, G. Ettel, T. Brandsberg, F. Brandsberg, M. Petitmermet,B. MuK ller, J. Mayer, E. Wintermantel

Department of Materials, Biocompatible Materials Science and Engineering, Swiss Federal Institute-Technology, ETH Zu( rich,Wagistrasse 23, CH-8952 Schlieren, Switzerland

Received 25 September 2000; accepted 13 February 2001

Abstract

A novel device for mechanical stimulation of primary adult rat bone marrow cells cultured on three-dimensional knitted textiles hasbeen prototyped. A method has been developed ensuring a well-de"ned, high-density, and reproducible cell seeding on the knittedfabric. After culturing for 18}52 days the cell}knit composites were subjected to uniaxial 2% stretching and relaxation. The frequencywas altered between 0.1Hz (196min, loading phase) and 0.01Hz (360min, resting phase). Identically treated knits without cellsexhibited a slight sti!ness reduction, whereas the sti!ness of knits with cells increased from cycle to cycle. The sti!ness increase wasfound to depend on the duration of the culture period before mechanical loading. Our data suggest that the extracellular matrixdeposited by the cells on the knit and intact microtubuli of living cells cause the observed sti!ness increase. In comparison to theunstrained static cell}knit composites cell proliferation and bone cell di!erentiation were reduced by the mechanical load. � 2001Elsevier Science Ltd. All rights reserved.

Keywords: Mechanical load; Culture; Bone marrow; knit; Sti!ness; Strain

1. Introduction

Mechanical loading governs the extent of growth anddi!erentiation of many tissues, including bone. Themechanical strain window and the e!ects on bone forma-tion and remodeling are determined by the frequency andthe magnitude of the strain [1,2]. The mechanism behindthe sequence from sensing the strain up to cell reaction isnot fully understood [3]. In vitro investigations are espe-cially predestinated to elucidate mechanisms of speci"cprocesses. Among others this knowledge is needed indesigning culture conditions for generating in vitro bonetissue. This engineered tissue can be potentially used toaugment bone or to "ll critical size bone defects.

In order to study e!ects of mechanical loading on bonemarrow cells a three-dimensional model system was de-veloped, which enables to stimulate cells mechanicallyunder well-de"ned conditions. Knitted textiles as poroussca!olds provide not only optimal spatial and nutritionalconditions for engineering of biological tissues, but arecapable of transferring mechanical load to cells. Further-more, knits have the advantage that cells can migrate tosites at which optimal nutritional and mechanical load-ing conditions are present [4]. Generally, cells are culti-vated as two-dimensional monolayers. Comparativestudies of behavior and performance of three- and two-dimensional cultured cells showed essential di!erences,with three-dimensional cultured cells behaving muchmore in vivo like [5}7]. The advantage of three-dimen-sional textile superstructures is that cells can be culturedand mechanically loaded in an in vivo like three-dimen-sional fashion. Therefore, a knit pulsating system wasdeveloped for uniaxial, cyclic stretching of knit}cell com-posites. The premise of the use of knits as sca!olds formechanical loading studies is that cells are homogene-ously covered by cells at a rather high cell density.

0142-9612/01/$ - see front matter � 2001 Elsevier Science Ltd. All rights reserved.PII: S 0 1 4 2 - 9 6 1 2 ( 0 1 ) 0 0 0 6 9 - 2

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556

57585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111

Fig. 1. Schematic representation of the knit used. The yarns are com-posed of 24 "laments drawn for reasons of simpli"cation as a singlethread with only one "lament. The knit is composed of 2 layers whichare alternating on the left and right side, changing sides betweenposition C and A (for the sake of clarity one yarn is drawn black).

Therefore, before starting the mechanical loading experi-ments, a cell seeding technique had to be developed thatenables a homogeneous high cell density coverage of theknits.

Under physiological conditions maximal strains ofabout 2000�� have been assessed in humans with a maxi-mal strain rate of about 40,000��/s [8]. The macroscopicstrain (�) is de"ned by the ratio of the induced increase inlength to the original length. Strains above 5000 �� arethought to be present in the pathological overload situ-ation. In these cases woven bone is formed similar toa bone repair reaction [9]. Strains above 10,000�� areassumed to occur in vivo after a bone fracture. However,only strains below 50,000�� are known to induce intra-membranous bone [10]. Di!erent e!ects of strains onosteoblast cell proliferation and di!erentiation in vitrohave been reported [3,11,12]. Usually, strains above4000�� are adequate to modify osteoblast-like cell func-tion, generally expressed by a change in cell shape. In thepresent study a situation was chosen to mimic wherein vivo intramembranous bone is formed. Therefore, theinvestigations described here use a macroscopic strainapplication of 20,000��.

The aims of the present study were: (i) to develop anoptimized knit pretreatment and cell seeding technique,(ii) to assess the di!erence in mechanical-properties-evolution of the knit}cell composite with and withoutcells as a result of mechanical stimulation, (iii) to revealdata to explain the latter obtained di!erence and (iv) todetermine the in#uence of mechanical loading on theproliferation and di!erentiation of bone cells in compari-son to the unstrained reference.

2. Materials and methods

2.1. Scawold preparation

The sca!old used (Fig. 1) was a knitted polyethylene-terephthalate (PET) multi"lament yarn. Each threadconsists of 24 highly textured "laments, having a dia-meter of 20 �m. The number of stitches per cmwas 15.2 inthe course and 15.5 in the wale direction. The thickness ofthe knit in relaxed state is 1.0mm. The size used was7.0 cm�4.0 cm. In order to optimize cell adhesion anda homogeneous cell distribution within the knits thee!ect of inclusion of three di!erent knit treatment stepswas tested. They consist of a cleaning step (step 1), a pre-stretching step (step 2) and a plasma treatment step (step3). They were performed one day before cell isolation.Regarding the cleaning step, knits were cleaned by sub-sequently bi-distilled water, sonicated for 120 s in acetoneand ethanol. In step 2, knits were prestretched by "xingthe knits on two opposite sides in a way that they couldbe stretched for 10% in the course direction. The knitswere autoclaved in this position afterwards. Step 3 repre-

sented a plasma activation of the knit surface. For this,knits were air plasma treated for 60 s at 2mbar, 10mHzand 60W (Tissue cleaner PDC-32G, Harrick, USA).After the plasma treatment the knits were transferredinto cell culture #asks (80 cm� substratum surface, Nunc,CH) containing 25ml �-MEM (Nunc, CH). After 30minthe medium was replaced by cell culture medium con-taining 89% �-MEM, 10% heat inactivated fetal calfserum and 1% antibiotics (PSN, Gibco, CH).

2.2. Preparation of the cell cultures

Rat bone marrow cells (RBMCs) were obtained fromtibiae and femora of male adult Wistar rats (250}300 g).The rats were killed by decapitation. The bones wereexcised, freed from soft tissue, without the cartilage endstransferred into isolation medium [13] and broken after-wards. RBMCs were dispersed by repeated pipetting.Pieces of bone, coagula and cell clusters were removed by"ltration over a 200�m mesh polyester (PET) "lter. Thecells were sedimented by centrifugation (14min at 30g)and resuspended in phosphate bu!ered saline (PBS).After recentrifugation pellets were resuspended in cellculture medium. The RBMC density was assessed usinga coulter counter (Coulter Z1, IG AG, ZuK rich) (countingwindow 7.5}21.7�m). Knits with 20�10� RBMC wereincubated in a cell culture incubator (5% CO

�, 95%

humidity) under gyratory shaking for 18}52 days. In the"rst experiment some knits were incubated without gyra-tory shaking. This period is termed preincubation period.For development and optimization of the cell seedingand knit pretreatment procedure, living cells on the knits

3170 A. Bruinink et al. / Biomaterials 22 (2001) 3169}3178

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556

57585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111

Fig. 2. Picture (A) and schematical representation of the mechanical loading device (B). The force}strain hysteresis of a spring (C) and the knits without(D) and with (E) cells at start and end of the mechanical loading experiments. 1 and 2: a knit stretched for 0% and 2%, respectively. a: location whereknits are "xed; b: part of the loading device which will be connected to the stretching motor; c: force sensor; d: cell}knit composite; e: culture dish withmedium; f: #exible metal membranes.

were stained by treating the cell}knit composite byadding thiazolylbluetetrazolium-bromide (MTT) to theculture medium. The staining intensity and distributionof the di!erently treated composites were compared bylight microscopy after incubating the composites undercell culture conditions for 1 h.

2.3. Testing conditions

To be able to mechanically stimulate and to followon-line mechanical properties of the cell}knit composites

after the preincubation period, knits were transferred intothe knit pulsating device (Fig. 2A and B). It consists of twolamellae made from PEEK between which the cell}knitcomposite is "xed. One of these lamellae is connected toa sensitive force sensor (quartz crystal sensor; Kistler In-strumente AG, CH), the other to a rotator responsible forthe stretching and relaxation of the cell}knit composite.The tops of the lamellae are connected to the moduleframe by #exible metal sheet membranes.

Of the knits the central 3 cm of the knit (course direc-tion) was used and the remaining parts were removed as

A. Bruinink et al. / Biomaterials 22 (2001) 3169}3178 3171

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556

57585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111

far as not needed for "xing into the device. The knits weremechanically loaded under cell incubator conditions. Inthe present experiments they were alternately stretchedby 2% (20,000�� or 20m�, simulating bone fractureconditions), in the course direction, sine wave, for196min at 0.1Hz or 2000��/s (termed loading phase)and for 360min at 0.01Hz or 200��/s (termed restingphase) (Fig. 2B). The mechanical strain of each single cellis dependent on the location within the knit ranging fromzero to above 50m� as a result of the complexity of theknit used. As concluded from light microscopy images ofknits loaded with and without 20m� strain, in the coursedirection maximal strains above 20m� were observed atD position in Fig. 1 measuring the distance betweenC and A, whereas at B, measuring the distance betweenA and C, strains of only 0}10m� could be observed.Simultaneously, the width of layers C and A was reducedby 0}7% corresponding to a strain of 0}70m�. Further-more, as a result of the elongation, a contraction of theknit perpendicular to the course direction, i.e., in the waledirection, was observed. The contraction was almost zeroat the position where the knit was connected to thedevice, but around 5% in the central part resulting in50m� strain.

To uncover the evolution of the knit sti!ness ande!ects on cell performance, identically treated knits with-out living cells were used for comparison. Furthermore,static knits with cells (not shifted, not stretched) and knitswith cells that were cyclic shifted instead of stretchedwere included in this study. The shifted knits were con-nected to the knit pulsating device only at one sideinstead of two sides, i.e., the knits were connected to theside without the force sensor. In the shifted knits it maybe assumed that the cells will predominantly be subjectedto #uid shear forces. Since latter cell}knit composites arenot connected to the force sensor, the sti!ness evolutioncannot be assessed in these composites.

The current study consists of 9 independent experi-ments of which 8 use knits with cells and oneknits without cells but preincubated with serumcontaining medium for 35 days. The knits with cells werepreincubated for 18 (3 experiments), 35, 38, 39, 45 and 52days. The cells on the 39-day preincubated knits werekilled by formalin vapor prior to the mechanical loadingexperiment. Serum incubated knits and knits with deadcells were included in the present study to discriminatebetween unspeci"c e!ects and e!ects induced by livingcells.

The prepared cell}knit composites were prestretchedwithin the mechanical loading device, giving rise toa spring constant of 0.33N/mm veri"ed by the piezoelec-tric force sensor. The well-de"ned starting conditionsguarantee reproducible mechanical stimulation. Sub-sequently, the mechanical stimulation was performed us-ing a total strain of 20m�. After the run-in period of 12 h,the force was in situ monitored continuously during each

loading phase for 76h (with the exception of one refer-ence measurement).

Fig. 2C represents the device calibration by the use ofa coil spring. The spring constant is given by the slope,and the negligible hysteresis reveals the elastic behaviorof the spring. Contrary to the spring, both the knit andthe cell}knit composite exhibit signi"cant hysteresis,which corresponds to the dissipation of mechanical en-ergy during one cycle of stimulation (compare Fig. 2Dand E). The sti!ness, as used in the present communica-tion, corresponds to the spring constant de"ned by theturning points 1 and 2.

To evaluate the importance of ECM in the sti!nessincrease of the cell}knit composite in one selected experi-ment a cell}knit composite (35 days preincubated) wastreated 3 days after start of the mechanical loading bya mixture of arabinofuranosylcytosine (ara-C, 0.1mM)and colchicine (0.1mM). A second cell}knit compositewas treated by a vehicle. Ara-C is known to kill prolif-erating cells [13], e.g., progenitor cells, "broblastic cellsand preosteoblasts, whereas colchicine inhibits micro-tubule polymerization. As tested in bone marrow cellcultures cultivated in 24 well plates and treated by thismixture for 2 and 3 days all proliferating cells died. Onlycells exhibiting osteoblast- and osteoclast-like morpho-logy were not a!ected. The mechanical loading and forcemeasurements proceeded for 56 h.

To determine the in#uence of mechanical loading onthe proliferation and di!erentiation of bone cells in com-parison to the unstrained reference, after termination of6 of the experiments, knits with living cells were cut intotwo identical pieces. One piece was used to measureMTT-conversion as an index for cell activity [13]. Theother piece was used to determine cell mass (total cellprotein) and DNA, alkaline phosphatase activity (ALP)(di!erentiation of progenitor cells into (pre-)osteoblasts)and tartrate acid phosphatase activity (TRAP) [14]. Ac-tive osteoclasts are known to synthesize high quantitiesof TRAP. Signi"cant e!ects (p(0.05) were determinedusing the ANOVA two factorial Dunn-Bonferroni test.Some representative knits were morphologically ana-lyzed using a scanning electron microscope.

3. Results

3.1. Optimization of knit pretreatment and seedingtechnique

Seeding under static conditions resulted in a high cellcoverage of the culture #ask substratum with nearly nocells present on the knits. In contrast, by culturing theknits under gyratory shaking conditions no cells at-tached to the #ask substratum but high cell densitieswere found on the knits. Without the knit cleaning andplasma treatment cells were present in clusters that wereirregularly distributed on the knit. By including both steps

3172 A. Bruinink et al. / Biomaterials 22 (2001) 3169}3178

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556

57585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111

Fig. 3. Scanning electronmicroscopic pictures of a knit without (A) andwith (B) cells after 18 days in culture. Please note that the biologicalmaterial (ECM with cells) is predominantly present at locations withhighest "lament density.

Fig. 4. Time evolution of force di!erence between relaxation and 2%stretching relative to start of the measurements of 9 independent experi-ments.

a homogeneous cell coverage of the knits was achieved atleast in the area of the knit used for mechanical loading.Inclusion of the plasma treatment step gives rise to in-creased cell densities as judged after 18 days in culture(Fig. 3). The inclusion of the prestretching step reducedsigni"cantly the sti!ness decrease of cell-free knits thatnormally occurs during mechanical loading when usingunstretched cell-free knits.

3.2. Ewects of mechanical loading on the stiwnessof the cell}knit composite

The sti!ness of the knit}cell composite increased dur-ing mechanical loading, whereas the sti!ness of knits

without cells was slightly reduced (Figs. 2 and 4). Theincrease in sti!ness was found to depend on cultureperiod before mechanical loading. Slight increase wasfound in cell}knit composites preincubated for only18 days. Maximal sti!ness increase was found forcell}knit composites preincubated for at least 35 days(Fig. 4). The observed sti!ness increase correlates withthe quantity of protein present on the knits, representingextracellular matrix (ECM) and cell protein, at the end ofthe mechanical loading experiment (Fig. 5A). Other bio-chemical parameters do not correlate with the sti!nessincrease although in the case of MTT-dehydrogenaseactivity (Fig. 5B) the mean value found for 38}52 prein-cubation cell}knit composites was 195% of the meanvalue found for 18 days preincubated composites. Re-garding cell di!erentiation, mean ALP and TRAP activ-ity values were virtually the same in cell}knit compositespreincubated for 18 or for 38}52 days.

As mentioned above the knits were stretched alternate-ly for 196min at 0.1Hz and for 360min at 0.01Hz. Inknits preincubated for 35}52 days the sti!ness increased

A. Bruinink et al. / Biomaterials 22 (2001) 3169}3178 3173

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556

57585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111

Fig. 5. Correlation between sti!ness increase and protein content (A) and MTT-formazan formation (B) at the end of the mechanical loadingexperiment in cell}knit composites that were preincubated before start of the mechanical loading for 18}52 days (mean experimental values$SD oftriplicate measurements are shown). The knit}cell composites shown were preincubated for 18 (�) or 35 days and more (�).

nearly exclusively during the loading phase. Such a fea-ture was not detected for 18 days preincubated knits. Forthese knits an almost continuous sti!ness increase wasobserved (Fig. 4).

By adding a mixture of ara-C and colchicine after3 days of mechanical loading of the treated knit about22% of the total cell number died as concluded from thereduction in MTT-dehydrogenase activity in treated ver-sus untreated composites at the end of the experiment. Inparallel the sti!ness increase was reduced by 70%. Thesti!ness of the untreated cell}knit reference compositeincreased further on with the same pattern as seen at theprevious 3 days.

3.3. Ewects of mechanical loading on cell morphologyand cell performance

Mechanical loading of the cell}knit composites result-ed in a change in cell plus ECM distribution as suggestedby scanning electron microscopy analysis (Fig. 6). Mech-anical loading resulted in a preferential alignment of theECM plus cells along the single "laments. Less bridgingof ECM plus cells between the "laments was observed inthe mechanical loaded knits. The scanning electronmicroscopy images, however, do not allow for di!erenti-ation between cells and ECM. In contrast, no di!erencesin cell distribution in strained and unstrained knits wereseen in cell}knit composites by light microscopy afterstaining cells with MTT.

Regarding cell function a signi"cant reduction in cellproliferation (DNA content) in stretched and non-

stretched but cyclically shifted (dislocated) knits in com-parison to static knits was found (23% and 31% reduc-tion, respectively). Also most other parameters weresigni"cantly reduced in comparison with the static knits(TRAP activity: 23% and 28%; ALP activity: 31% and17%; cell protein: 7% and 20%; MTT conversion: 1%(not signi"cant) and 16%, respectively) (Fig. 7). The cellson stretched and those cyclically shifted but unstretchedknits behaved similarly. No correlation seems to existbetween the mechanical load induced change in cell per-formance (as de"ned by DNA and cell protein content,ALP and TRAP activity) and the length of the cultureperiod before loading.

4. Discussion

In the present study, the e!ects of mechanical stimula-tion of cells cultured in a three-dimensional organizedcell}knit composite were investigated, using primarylong bone marrow cells of adult rats. The main para-meters of the study were cell}knit composite sti!ness andcell function. This is the "rst study where cells are cul-tured on a knitted fabric and thereafter subjected tomechanical load.

Since the di!erentiation of bone marrow cells contain-ing mainly non-anchorage dependent progenitor cells toanchorage dependent cells takes a few days, with currenttechniques it was impossible to seed knitted sca!oldsreproducibly with bone marrow cells. Xiao andcoworkers found that in contrast to static conditions

3174 A. Bruinink et al. / Biomaterials 22 (2001) 3169}3178

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556

57585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111

Fig. 6. E!ects of mechanical loading on rat bone marrow cell morpho-logy. Scanning electronmicroscopy images of a static knit (A) anda loaded knit with cells at the end of the mechanical loading experiment(B). Note that in (B) the biological material (ECM plus cells) is predomi-nantly aligned along the "laments. The knit}cell composites shownwere preincubated for 18 days.

dynamic seeding of "broblasts onto porous sca!oldsleads to e$cient as well as homogeneous cell attachment[15]. A homogeneous cell coverage is a prerequisite tosuccessfully perform reproducible mechanical loading ex-periments. In agreement we found that nearly no cellsattached to the static knits. However, by cultivating theknits under gyratory shaking conditions, cells were notable to attach to the #ask surface anymore but to theknit. By including the cleaning and plasma treatmentsteps a homogeneous high-cell-density coverage of theknits was achieved. The cleaning step removes "lamentsizing, which may reduce serum protein adsorption,

whereas plasma treatment increases surface charge. Theinclusion of the prestretching step reduced signi"cantlythe change in the hysteresis "gure of cell-free knits duringmechanical loading. Thus by including all three knitpretreatment steps and by culturing the cell}knit com-posites under gyratory shaking conditions optimal pre-requisites were revealed for starting the mechanicalloading experiments.

One of the main goals of the present study was toassess the di!erence in mechanical-properties-evolutionof the knit}cell composite with and without cells asa result of mechanical stimulation and to reveal data toexplain the latter obtained di!erence. Our study demon-strates that during mechanical loading cell}knit com-posite sti!ness is only increased if living cells are presenton the knit (Fig. 4). This increase in sti!ness including itsevolution with time was found to depend on the cultureperiod before the cell}knit composites were transferredto the loading device. The increase in sti!ness may be dueto (i) the release of ECM components, (ii) the increase incell mass, (iii) tensional reactions of the cells togetherwith the increase in cell number. (iv) All of these maygreatly be a!ected by the cell type composition and thedegree of di!erentiation of the cells present on the knit.

Cell diwerentiation: Since ALP and TRAP activities (asindex for osteoblast and osteoclast di!erentiation, re-spectively) were similar in the various mechanical loadedcell}knit composites, one may conclude that the ob-served di!erences in sti!ness increase cannot be tracedback to an altered degree in cell di!erentiation at least onthe osteoblast and osteoclast level.

ECM and cell number: Of the parameters tested onlythe protein content of the cell}knit composite correlateswith the extent of sti!ness increase. The proteins presenton the knits represent ECM and cell proteins. The totalquantity of cell protein is determined by cell size and cellnumber. Since no relationship exists between cell activity(MTT-dehydrogenase activity as an index for active liv-ing cell number) and increase in sti!ness, it is hy-pothesized that the increase in ECM quantity ispredominantly responsible for the observed sti!ness in-crease.

Tensional reactions: The importance of ECM in theobserved sti!ness increase is partly put in question bya preliminary experiment where a cell}knit compositewas treated 3 days after start of the mechanical loadingby a mixture of ara-C, killing proliferating cells, and ofcolchicine, destroying microtubuli. Since the reduction insti!ness surpasses the decrease in the number of livingcells taking MTT-dehydrogenase activity as an index byat least 3.2-fold we hypothesize that an important part ofthe sti!ness increase besides that by the ECM is based onintact microtubuli of living cells and not on cell numberper se. Regarding tensional cellular reactions, it is knownthat cyclic mechanical loading of "broblasts results ina modi"cation in cell orientation and cell mediated

A. Bruinink et al. / Biomaterials 22 (2001) 3169}3178 3175

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556

57585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111

Fig. 7. E!ects of mechanical loading on TRAP (A) and ALP (B) activities, cell protein (C) and DNA (D) content, and on MTT-dehydrogenase activity(E) using cell}knit composites that were cultured for di!erent periods before stimulation (mean$SD of triplicate measurement of each singleexperiment).

contractile tension [16}18]. Whereas micro"laments arerelated to the mechanism of signal transduction in re-sponse to mechanical stimulation [11], microtubules area potential target for translating changes in externallyapplied mechanical stimuli to alterations in cellularphenotype [19], e.g. cell shape and orientation. Thus, ourresults suggest that ECM and intact microtubuli of livingcells also play a key role in the observed sti!ness increase.

In summary, the increase in sti!ness is not based ona single but on multiple factors including ECM and celltensional forces.

Under the mechanical strain regimen chosen resultsindicate that a slight mechanical loading (dislocation ofthe knit without active stretching) a!ects cell perfor-mance, e.g., proliferation and di!erentiation. These ef-fects are found to be comparable to those of stretchedknits although regarding DNA, protein content andTRAP activity these e!ects were more pronounced in the20m� strained knits. In analogy, modi"cation of cellperformance has been reported for in vitro studies usingosteoclast- or osteoblast-like cells that are slight mechan-ical loaded with strains within the range of the overuse

3176 A. Bruinink et al. / Biomaterials 22 (2001) 3169}3178

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556

57585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111

window de"ned as 2}5m� or by strains between 10 and170m� being within the pathological overload window[20}24]. In knits that were only shifted the cell stretchingwas minimal, whereas the #uid shear forces were compa-rable to those that were stretched. Thus #uid shear forcesseem to be mainly responsible for the induced modi"ca-tion in cell performance. Several research groups sugges-ted that #uid #ow is the primary stimulus by which bonecells respond to mechanical strain [25, 26].

In line with the study of Matsuda and coworkers [27]we found that cell proliferation was diminished com-pared to the static control. In analogy to our studyMatsuda and coworkers stimulated the cells at a fre-quency of 0.1Hz and a strain within the overuse window(90m�). Most studies describing no e!ect have been per-formed at frequencies between 0.1 and 0.25Hz [28], andthose describing a positive e!ect on cell proliferationhave been performed at frequencies between 0.25 and1Hz [29]. In analogy to these in vitro "ndings in vivostimulation of non-fractured rat tibiae resulted in in-crease in bone formation at frequencies at and above0.5Hz but not at 0.2Hz and below [26]. So far it is notknown, which part of the stimulus, e.g., frequency, dura-tion, and rate of the applied stimulus, is decisive for theinduction of cellular responses. It has been reported thatfollowing mechanical loading at a frequency of around0.1Hz osteoblasts react with cell membrane depolariz-ation whereas at frequencies above 0.2Hz cell membraneis hyperpolarized [30]. Thus, the induced membranepotential (hyperpolarization or depolarization) may playa crucial role in the type of cell response. In line with thishypothesis, electric "eld stimulation of semi-monolayerosteoblastic cell cultures known to reduce transmem-brane potential has been described to reduce cell prolifer-ation [31]. Additional experiments performed at di!erentmechanical loading frequencies are needed to verify thecurrent hypothesis.

Besides a reduction in cell proliferation we also detec-ted a reduction in di!erentiation. This "nding is in linewith other studies of mechanical stimulation of osteo-clast-like cells [24] but is in contrast to those usingosteoblast-like cells. Regarding ostoblast-like cells, gen-erally, no, or cell di!erentiation promoting, e!ects werereported [20,22,28,29]. In addition to di!erences instimulation regimen, this e!ect may be explained by thetype of sca!old ("laments versus plain membrane surfa-ces) and the dimensionality of the culture used for mech-anical stimulation. It is generally accepted thatmechanical stimulation is transduced into a cellular re-sponse by induction of various biochemical signals suchas growth factors that regulate further cellular reactions.These signals may promote an inhibition of cellproliferation. Generally, two-dimensional cultureson a #at, membrane-like surface have been usedfor mechanical stimulation. In this environment sig-naling factors that are released, are diluted before

contacting neighboring cells compared to cells that areorganized in close contact with each other in a three-dimensional fashion.

In conclusion, optimal conditions for knit pretreat-ment and cell seeding were developed. It is shown thatduring cyclic uniaxial mechanical loading of a three-dimensional knit}bone marrow cell composite the sti!-ness of the knit}cell composite is increased. Evidence wasfound that ECM plays a key role in the observed sti!nessincrease. As a result of the mechanical loading cell func-tion, regarding cell proliferation, protein synthesis, ALPand TRAP activity, of cells cultured on these knits arereduced.

References

[1] Turner CH, Forwood MR, Otter MW. Mechanotransduction inbone: do bone cells act as sensors of #uid #ow?. FASEBJ 1994;8:875}8.

[2] Turner CH, Forwood MR, Rho JY, Yoshikawa T. Mechanicalloading thresholds for lamellar and woven bone formation.J Bone Miner Res 1994;9:87}97.

[3] Neidlinger-Wilke C, Wilke HJ, Claes L. Cyclic stretching ofhuman osteoblasts a!ects proliferation and metabolism: a newexperimental method and its application. J Orthop Res1994;12:70}8.

[4] Vandenburgh HH. A computerized mechanical cell stimulator fortissue culture e!ects on skeletal muscle organogenesis. In VitroCell Dev Biol 1988;24:609}19.

[5] Ho!man RM. To do tissue culture in two or three dimensions?That is the question. Stem Cells 1993;11:105}11.

[6] Benya PD, Sha!er JD. Dedi!erentiated chondrocytes reexpressthe di!erentiated collagen phenotype when cultured in agarosegels. Cell 1982;30:215}24.

[7] Liu M, Xu J, Souza P, Tanswell B, Post M. The e!ect of mechan-ical strain on fetal rat lung cell proliferation: comparison of two-and three dimensional culture systems. In Vitro Cell Dev BiolAnim 1995;31:858}66.

[8] Burr DB, Milgrom C, Fyhrie D, Forwood M, Nyska M, Fi-nestone A, Hoshaw S, Saiag E, Simkin A. In vivo measurement ofhuman tibia strains during vigorous activity. Bone1996;18:405}10.

[9] Duncan RL, Turner CH. Mechanotransduction and the func-tional response of bone to mechanical strain. Calcif Tissue Int1995;57:344}58.

[10] Claes LE, Heigele CA, Neidlinger-Wilke C, Kaspar D, Seidl W,Margevicius KJ, Augat P. E!ects of mechanical factors on thehealing process. Clin Orthop Relat Res 1998;355S:S132}47.

[11] Toma CD, Ashkar S, Gray ML, Scha!er JL, Gerstenfeld LC.Signal transduction of mechanical stimuli is dependent on micro-"lament integrity: identi"cation of osteopontin as a mechan-ically induced gene in osteoblasts. J Bone Miner Res1997;12:1626}36.

[12] Jones DB, Nolte H, Scholubbers JG, Turner E, Veltel D. Bio-chemical signal transduction of mechanical strain in osteoblast-like cells. Biomaterials 1991;12:101}10.

[13] Bruinink A. Serum-free monolayer cultures of fetal chick brainand retina: immunoassays of developmental markers, mathemat-ical data analysis, and establishment of optimal culture condi-tions. In: Zbinden G, editor. The brain in bits and pieces.Zollikon: M.T.C. Verlag, 1992. p. 23}50.

[14] Schroeder A, Francz G, Bruinink A, Hauert R, Mayer J, Winter-mantel E. Titanium containing amorphous hydrogenated carbon

A. Bruinink et al. / Biomaterials 22 (2001) 3169}3178 3177

1234567891011121314151617181920212223242526

27282930313233343536373839404142434445464748495051

"lms (a-C :H/Ti): surface analysis and evaluation of cellular reac-tions using bone marrow cell cultures in vitro. Biomaterials2000;21:449}56.

[15] Xiao Y-L, Riesle J, Van Blitterswijk CA. Static and dynamic"broblast seeding and cultivation in porous PEO/PBT sca!olds.J Mater Sci Med 1999;10:773}7.

[16] Eastwood M, Mudera VC, McGrouther. E!ect of precisemechanical loading on "broblast populated collagenlattices: morphological changes. Cell Motil Cytoskeleton1998;40:12}21.

[17] Brown RA, Prajapati R, McGrouther DA, Yannas IV, EastwoodM. Tensional homeostasis in dermal "broblasts: mechanical re-sponses to mechanical loading in three-dimensional substrates.J Cell Physiol 1998;175:323}32.

[18] Thoumine O, Ott A. Time scale dependent viscoelastic and con-tratile regimes in "broblasts probed by microplate manipulation.J Cell Res 1997;110:2109}16.

[19] Putnam AJ, Cunnigham JJ, Dennis RG, Linderman JJ, MooneyDJ. Microtubule assembly is regulated by externally appliedstrain in cultured smooth muscle cells. J Cell Sci 1998;111:3379}87.

[20] Nihioka S, Fukuda K, Tanaka S. Cyclic stretch increases alkalinephophatase activity of osteoblast-like cells: a role for prosta-glandine E2. Bone Miner 1993;21:141}50.

[21] Damien E, Price JS, Lanyon LE. The estrogen receptor's involve-ment in osteoblast's adaptive response to mechnaical strain.J Bone Miner Res 1998;13:1275}82.

[22] Neidlinger-Wilke C, Stalla I, Claes L, Brand R, Hoellen I,RuK benacker S, Arand M, Kinzl L. Human osteoblasts fromyounger normal and osteoporotic donors show di!erences inproliferation and TGF-�-release in response to cyclic strain. J Bio-mech 1995;28:1411}8.

[23] Meyer U, Meyer T, Jones DB. No mechanical role for vinculin instrain transduction in primary bovine osteoblasts. Biochem CellBiol 1997;75:81}7.

[24] Rubin J, Fan X, Biskobing DM, Taylor WR, Rubin CT. Osteoc-lastogenesis is repressed by mechanical strain in an in vitro model.J Orthop Res 1999;17:639}45.

[25] Smalt R, Mitchell FT, Howard RL, Chambers TJ. Mechanotran-duction in bone cells: induction of nitric oxide and prostaglandinsynthesis by #uid shear stress, but not by mechanical strain. In:Sinzinger H, Samuelsson B, Vane JR, et al., editors. Recentadvances in prostaglandin, thromboxane, and leukotreine re-search. New York: Plenum Press, 1998. p. 311}4.

[26] Turner CH, Forwood MR, Otter MW. Mechanotransdiction inbone: do bone cells act as sensors of #uid #ow? FASABJ 1994;8:875}8.

[27] Matsuda N, Morita N, Matsuda K, Watanabe M. Proliferationand di!erentiation of human osteoblastic cells associated withdi!erntial activation of MAP kinases in response to epidermalgrowth factor, hypoxia, and mechanical stress in vitro. BiochemBiophys Res Commun 1990;249:350}4.

[28] Rubin J, Murphy T, Nanes MS, Fan X. Mechanical strain inhibitsexpression of osteoclast di!erentiation factor by murine stromalcells. Am J Physiol Cell Physiol 2000;278:C1126}32.

[29] Thomas GP, El Haj AJ. Bone marrow stromal cells are loadresponsive in vitro. Calcif Tissue Int 1996;58:101}8.

[30] Salter DM, Robb JE, WrightMO. Electrophysiological responsesof human bone cells to mechanical stimulation: evidence forspeci"c integrin function in mechanotransduction. J Bone MinerRes 1997;12:1133}41.

[31] McLeod KJ, Donahue HJ, Levin PE, Fontaine MA, Rubin CT.Electric "elds modulate bone cell function in a density-dependentmanner. J Bone Miner Res 1993;8:977}84.

3178 A. Bruinink et al. / Biomaterials 22 (2001) 3169}3178